Many adolescents and young adults engage in addictive behaviors, such as alcohol, tobacco, and cannabis use [1] or problematic internet use [2]. Vocational schools are promising settings for prevention efforts among adolescents and young adults [3]. First, vocational school students report even higher rates of substance use than their peers [4,5]. Second, these behaviors often tend to cluster among vocational students, making them more vulnerable to the development of noncommunicable diseases [5,6]. Representative data from German vocational students in Mecklenburg-Western Pomerania showed that 57% of vocational students had 2 or more problematic use patterns related to alcohol, tobacco smoking, cannabis, internet use, gaming, or gambling [5]. The highest clustering patterns among vocational students were found for tobacco smoking and alcohol use [5,6], and for tobacco and cannabis [5]. The clustering of addictive behavior is higher among male vocational students, those who are younger, those with lower educational attainment, and those in vocational preparation or production schools [5]. Nevertheless, vocational students differ widely in their risk profiles, that is, the combination and number of addictive behaviors. Therefore, flexible interventions are needed that can be tailored to different risk profiles.

Digital interventions have been successfully used to prevent and reduce addictive behaviors such as substance use, among adolescents and young adults [7-9], allowing for automated delivery of highly individualized content. In particular, mobile interventions have been shown to be effective in reducing multiple addictive behaviors among vocational students [10,11].

While results regarding efficacy are encouraging, consistent intervention usage remains a challenge for digital interventions to reduce substance use in general [12,13] as well as for adolescents and young adults [8,9]. Low rates of module completion and intervention revisits are common. For example, a recent review of digital interventions to reduce substance use among adults found that on average, 60% of modules were completed, and on average, 47% of participants completed all modules [13]. In their 8-week web-based health promotion intervention for vocational students, Stassen et al [14] found that only 16.6% of all potential users logged in at least once, of which 57.4% revisited the platform. In their app-based addictive behavior intervention for vocational students, Pietsch et al [10] addressed tobacco, e-cigarettes, alcohol, and cannabis use as well as gambling and digital media-related behaviors using a voluntary commitment approach to reduce or abstain from one of these behaviors. However, only half of the students in the intervention group used the app and started a 2-week abstinence challenge.

Interventions that address multiple behaviors may add another layer of complexity. When individuals engage in multiple addictive behaviors, addressing them all at once may be overwhelming [15], potentially leading to lower usage. Previous research suggests that focusing on 2 behaviors may be optimal in terms of adherence to component recommendations [16] and intervention efficacy [15]. Additionally, completion rates are likely to differ across delivered behavior modules [17-19]. For example, in their digital lifestyle intervention, Schulz et al [17] found that completion of the smoking module was 26% compared with 47% for the alcohol module.

Low usage is often linked to low exposure of participants to the intervention content, which in turn affects the effects seen [20,21]. Identifying predictors of usage could be used to determine specific aspects that contribute to better exposure to intervention content. A few reviews have been published on potential determinants of usage in digital interventions to reduce substance use in adult populations [12,13,22]. Intervention features that have been highlighted as important for increased usage include tailoring, reminders, customized content and features, and gamification or rewards. These were consistent with findings from a review of digital health interventions for adolescents [23]. In terms of user characteristics, the review by Jakob et al [12] found that female gender, lower substance use, and higher education were associated with higher usage in adults. Age was also associated with usage, but varied by substance: older age was associated with higher usage in digital alcohol interventions, whereas younger age was associated with higher usage in digital tobacco interventions.

To date, few studies have examined user characteristics as potential determinants of usage in digital substance use interventions among adolescents or young adults. In the mobile phone-based smoking cessation intervention for Swiss secondary and vocational school students by Paz Castro et al [24], those with a stable usage trajectory were younger, more likely to have a nonimmigrant background, and reported more perceived benefits of smoking cessation and binge drinking at baseline. Similarly, a randomized controlled trial of a mobile phone-based life-skills intervention for addiction prevention among Swiss secondary school students showed higher usage among those with lower alcohol consumption and those from upper secondary schools [25]. In addition, younger age, nonimmigrant background, and medium versus low levels of stress were associated with higher usage in that intervention in noncontrolled settings [26].

The “ready4life” app, initially developed and evaluated in Switzerland [27], aims to prevent and reduce multiple addictive behaviors among vocational students and offers 6 behavior modules: alcohol, tobacco, cannabis, social media and gaming, stress, and social competencies. After receiving feedback on their individual risks and competencies in the form of a traffic light system to guide choice, users can freely choose 2 of the 6 modules to be coached in for a total of 16 weeks. Because factors such as low social competence and high work stress are likely to contribute to addictive behaviors [28,29], these topics are included as modules. We have previously shown that students have high adherence to module recommendations [30]. Usage of the “ready4life” app was stimulated by tailored content, push notifications, social and gamification features (quizzes, contests, earning credit points based on completing the weekly dialogue, or winning prizes), and personal support in the form of an “ask-the-expert” function [12]. There were weekly chats as well as self-initiated dialogues that were accessible to users at any time.

An earlier pre-post study of 5896 Swiss vocational school students provided initial insights into how students used “ready4life” [31]. At the end of the first module in week 8, only 10%‐19% still participated in the weekly chat. Self-initiated dialogues were not used at all by 61% of the students. Longer program use was observed for females and those who used more self-initiated dialogues. Other variables, such as age, self-efficacy, educational track, starting month, and delivered behavior module did not show an association with usage. “ready4life” was recently evaluated in 2 cluster randomized controlled trials among 1351 Swiss [11] and 2545 German vocational school students [32], which demonstrated its feasibility and effectiveness. Intention-to-treat comparisons showed significant positive effects of the intervention on problematic alcohol use and internet use over 6 months in the Swiss study and on social competence, stress, problematic internet use, and tobacco consumption over 12 months in the German study. However, according to the Swiss study [11], usage was rather low. On average, students in the intervention group completed 2.1 (SD 3.5) of the 16 possible weekly dialogues, with 41% completing no dialogues, and 39% completing 1 or 2 dialogues.

In summary, research on the usage of digital interventions remains scarce for both (1) vocational school students and (2) interventions targeting multiple behaviors. Specifically, only 1 study [31] has examined the determinants of intervention usage among vocational students, and there is currently no data from German samples. Given the increased vulnerability of vocational students to clustered addictive behaviors, understanding factors that enhance usage is essential.

This study aims to add to the existing knowledge by investigating the usage of “ready4life” based on data from the German evaluation study [32,33]. Specifically, we aim to analyze student characteristics as potential determinants of usage. Based on the previous literature, usage was expected to be in the low to moderate range and to be associated with various student characteristics, such as gender.

Sociodemographics (ie, gender, age) and individual prevention needs (ie, addictive behaviors, social competence, and perceived stress) were considered as potential determinants of usage based on their relevance in previous research [12,24-26,31]. As the prevalence of addictive behaviors varies according to the educational track of vocational students [4,5], and attrition rates are likely to differ between academic and nonacademic tracks [34], we also analyzed educational track as a potential determinant of intervention usage. Self-efficacy was included because of its role in motivation and sustained effort in the behavior change process [35].

This is a secondary analysis of data from a cluster randomized controlled trial among German vocational students testing the efficacy of the app-based addiction prevention program “ready4life.” The protocol and main results have been published elsewhere [32,33]. The trial was registered in the German Clinical Trials Register: DRKS00022328.

Details on the flow and characteristics of the participants can be found in Guertler et al [36]. In Germany, vocational schools are an integral part of the educational landscape, offering a variety of programs that lead to different educational and career paths. While many students participate in vocational training, which combines part-time vocational classroom instruction with practical on-the-job training, vocational schools also offer vocational preparation classes and vocational grammar school classes that allow students to earn a university entrance certificate. The sample of this study consisted of vocational students from 5 German federal states who were enrolled in vocational training, preparation, or grammar school. A total of 376 classes from 35 schools were randomized to intervention (n=186) or control (n=190) groups to avoid contamination of conditions. Between October 2020 and March 2022, classes were introduced to this study during school hours and students were invited to download the “ready4life” app. After downloading the app, all students participated in an anonymous in-app screening on their prevention needs related to alcohol, tobacco, and cannabis consumption, internet use, social competencies, and stress. Students then provided digital informed consent and contact details (email or phone number). Vocational students, aged 16+ years with smartphones who provided contact information for follow-up data collection were eligible to participate in this study. A total of 4225 app downloads were recorded, and 2568 students provided informed consent. The participation rate was 46.7% (2508/5370) among students aged ≥16 years who had the correct app version and belonged to classes where the number of students present during the introduction was known. All study participants then received in-app feedback on their individual risks and competencies in the form of a traffic light system for each behavior assessed by the in-app screening.

Students in the intervention group received 16 weeks of coaching via the “ready4life” app, as described below. The control group received only a link to information on improving health behaviors and could access the coaching after 12 months. Both study groups were invited to web-based follow-up chat sessions at 6 and 12 months via text message or email. At the end of the study, prizes were raffled, stratified by study group. In the original study [32], 1286 students from the intervention classes and 1282 from the control classes participated. In the present study, only the participants assigned to the intervention group were examined (Figure S1 in Multimedia Appendix 1).

Details on the development and content of the intervention can be found in another publication [33]. “ready4life” aims to prevent or reduce addictive behaviors and promote life skills among vocational students. Intervention participants could freely choose 2 of 6 available behavior modules: alcohol, tobacco, cannabis, social media and gaming, stress, and social competencies (Figure 1). The 2 chosen modules were presented in random order. Each module included 8 weeks of individually tailored coaching. The coaching involved weekly 5-minute chat sessions initiated by the conversational agent. The module-specific content and goals for each week can be found in the study by Schmidt et al [33]. Additionally, self-initiated chat dialogues could be started at any time after the weekly chat was completed. For each module, 4 to 5 self-initiated chats were available (Table S1 in Multimedia Appendix 1), which could be repeated as often as desired.

Chats included media (videos, images, or links), quizzes, contests, challenges, and weekly push notifications. Contests were conducted during the third week of each module. Students were encouraged to upload texts or photos showing, for example, how they have fun without a smartphone or computer (social media and gaming module) or what relaxes them when they are stressed (stress module). Uploaded content was reviewed for appropriateness by study staff. The top 3 photos with the most likes from the community were displayed. In the sixth week of each module, personal challenges were set. Students could choose 1 of several available challenges at the beginning of that week. For example, in the tobacco module, students could choose not to smoke for a while, observe smoking behavior, or help friends quit smoking. At the end of the week, the conversational agent contacted the students to ask them about their experience with the challenge. In the second week of the tobacco and alcohol module, participants could opt in to additional push notifications to receive daily tips to prepare for and support smoking cessation or low-risk alcohol consumption.

In week 5, an “ask-the-expert” feature was activated within the main menu. For 2 weeks, students could submit their questions, which were answered by a professional (eg, psychologist). The most frequently asked questions and answers were made available to everyone anonymously. Students who indicated a need for professional intervention (eg, mentioning suicidal ideation or other crises) were provided with contact information for a local, free, 24/7 telephone counseling service.

Additionally, to increase usage, participants were able to earn credit points by completing the weekly dialogue. This increased their chances of winning prizes.

Ethical approval was granted by the ethics committees of the University of Lübeck (number 19‐419) and the University medicine Greifswald (BB 024/20). Digital informed consent was obtained from vocational students via the app, with no additional parental consent required under the EU General Data Protection Regulation. However, underage participants were advised to inform their parents or guardians. Students completed the app-based screening anonymously. For those who provided informed consent, research data were pseudonymized, with contact details (email or phone number) stored separately. Names were not collected; students chose nicknames. App data has been encrypted to ensure data privacy. Participants were not offered any compensation other than the chance to win prizes.

The sample consisted of 1266 students from the intervention group who received their assigned intervention (females n=569, 44.9%; mean age 19.53, SD 3.57 years; Table 1). Most of the students were in vocational training (n=786, 62.1%) and in their first or second year of education (n=890, 70.3%). On average, students engaged in 2.08 (SD 1.12) of 4 addictive behaviors (Table 1).

Table S3 in Multimedia Appendix 1 summarizes the usage across the sample. On average, participants started 4.8 (SD 5.1) and completed 4.3 (SD 5.2) weekly in-app chats out of the 16 possible. The median number of weekly in-app chats started was 2 (IQR 1‐7), and the median number of weekly in-app chats completed was 2 (IQR 1‐6). A total of 68/1266 (5.4%) did not start, and 264/1266 (20.9%) did not complete any of the weekly chats. However, 331/1266 (26.2%) completed one, 145/1266 (11.5%) two, 89/1266 (7%) three, 172/1266 (13.6%) four to eight, and 265/1266 (20.9%) nine or more of the weekly chats. Figure 2 shows the percentage of students who did not start, started, and completed the chat for each week separately. As can be seen, the largest drop in participation occurred between the first and second week and between the second and third week. Depending on the module combination chosen, a number of 8 to 10 self-initiated chats were available (Table S1 in Multimedia Appendix 1). Most students (903/1266, 71.3%) completed none of the self-initiated chats, 194/1266 (15.3%) completed one, and 169/1266 (13.3%) completed two or more (median 0, IQR 0‐1; mean 0.6, SD 1.2). Figure 3 illustrates usage time in terms of the percentage of students still being active over time. At the start of week 3, a total of 49.8% (630/1266) of the students were still using the app, while 50.2% (636/1266) had stopped using the app. By week 16, only 15.4% (195/1266) were still using the app. Accordingly, the median for the last weekly chat activity was 2 (IQR 1‐10) weeks, with a mean of 5.6 (SD 5.8) weeks. Usage of the first module (weeks 1 to 8) was generally higher than that of the second module (weeks 9 to 16; Table S3 in Multimedia Appendix 1).

Of the intervention group participants who received their assigned intervention, 30/1266 (2.4%) did not make a module choice. Of those who made a choice, the most commonly chosen module was stress (818/1236, 66.2%), followed by social media and gaming (625/1236, 50.6%), alcohol (360/1236, 29.1%), social competencies (306/1236, 24.8%), tobacco (232/1236, 18.8%), and cannabis (131/1236, 10.6%).

Usage during the first 8 weeks was numerically highest for the social competence module (Table S4 in Multimedia Appendix 1). Using the social competence module as a reference in multilevel negative binomial regressions, the number of started and completed weekly chats, as well as usage time, were significantly lower for the social media and gaming, alcohol, and tobacco modules, with no significant difference for the stress and cannabis modules. The results for started and completed self-initiated chats were in the same direction, but only social media and gaming and tobacco were significantly different from the social competence module (Table S5 in Multimedia Appendix 1).

Table S6 in Multimedia Appendix 1 shows the unadjusted association of student characteristics with intervention usage. The intraclass correlations for the usage parameters ranged from 6.9% to 11.5%, indicating a modest influence of class membership on usage.

A significantly higher number of completed weekly dialogues was observed for females (P<.001; incidence rate ratio [IRR] 1.55, 95% CI 1.33‐1.80), while fewer were completed by individuals with higher self-efficacy (P=.04; IRR 0.96, 95% CI 0.93‐0.998), higher social competencies (P<.001; IRR 0.97, 95% CI 0.95‐0.98), and individuals engaging in more addictive behaviors (P<.001; IRR 0.87, 95% CI 0.82‐0.93). Concerning educational track, service and sales workers (P=.049; IRR 0.70, 95% CI 0.49‐0.999), and craft-related trades workers or plant and machine operators and assemblers (P<.001; IRR 0.57, 95% CI 0.41‐0.80) showed a lower number of completed weekly dialogues compared to professionals, technicians, and associate professionals. Clerical support workers (P=.16; IRR 0.76, 95% CI 0.52‐1.11), vocational grammar school students (P=.88; IRR 0.98, 95% CI 0.72‐1.33), and vocational preparation (P=.07; IRR 0.71, 95% CI 0.49‐1.03) did not differ significantly from professionals, technicians, and associate professionals in the number of completed weekly dialogues. Predictors of completed self-initiated chats and usage time largely aligned with the findings for weekly dialogues. Additionally, those with higher perceived stress completed significantly more self-initiated chats (P<.001; IRR 1.19, 95% CI 1.08‐1.31). Age and year of education were not significantly associated with any of the usage parameters.

The associations of social competence and the number of addictive behaviors with usage measures were independent of gender, age, and year of education (Table S7 in Multimedia Appendix 1). Self-efficacy showed a robust association with the number of self-initiated chats completed, while perceived stress showed no significant associations with any usage measure when adjusted for gender, age, and year of education. Associations between educational tracks and usage also remained, but were smaller and no longer statistically significant across all usage measures.

This study examined how vocational school students used the app-based addiction prevention program “ready4life.” The main findings were: (1) relatively low usage, with an average of 4 of 16 (SD 5.2) weekly chats completed, and most (903/1266, 71.3%) students did not complete any self-initiated chats. About half of the students stopped using the app before the start of week three. (2) Higher usage was observed among females, individuals with lower self-efficacy, lower social competence, higher perceived stress, and fewer addictive behaviors. Additionally, professionals, technicians, and associate professionals, as well as students in vocational grammar school or vocational preparation, showed higher usage than students in training for operational and support roles.

The strengths of this study include (1) targeting vocational students, an underserved group in digital intervention research, particularly in terms of usage analyses; (2) being the first study to examine determinants of usage in a multi-behavioral digital intervention for German vocational students; (3) conducting detailed analyses, including module-specific usage, multiple usage metrics, and module sequencing effects; and (4) using a facilitated access approach, which provides a more representative sample than web-based or media recruitment methods [50].

The results of this study should be considered in light of its limitations. First, as a secondary analysis, the original study was not designed to address the research questions posed in this paper. Second, although previous research has suggested that usage is often associated with efficacy [20,21], we have not yet tested this association in our study [51]. Third, usage was analyzed based on 3 parameters: weekly chats, self-initiated chats, and usage time. Other components, such as media content viewed, ask-the-expert, and number of logins, were not included in this analysis because the data were not automatically stored. Fourth, completion of weekly chats was rewarded with credits (and credits were tied to a prize draw), which may have influenced students’ intervention usage. Although we did not experimentally test this assumption, previous reviews have suggested this relationship (eg, [12]). Furthermore, in the focus group conducted at the end of our study, students cited the increased chances of winning prizes as a reason for continued use [52]. Fifth, findings may not be generalizable to other adolescent populations. Additionally, our previous analyses [36] showed that participants in this study differed from nonparticipants on several baseline variables, with students from nonprofessional backgrounds (eg, service and sales workers, or vocational preparation) and males being underrepresented among study participants. Conversely, those with higher support needs were overrepresented among this study’s participants, including students with lower social competence, lifetime cannabis use, higher problematic internet use, and higher perceived stress. Sixth, data collection during the COVID-19 pandemic may have influenced addictive behaviors or usage. Seventh, there may be other variables influencing usage that we have not analyzed. For example, previous studies of digital substance use interventions among adolescents or young adults found that nonimmigrant backgrounds and perceived benefits of smoking cessation [24,26] were associated with higher usage. Finally, reliance on self-report data carries the risk that the results may have been influenced by social desirability.

Our study adds to the existing evidence that maintaining consistent use in adolescents and young adults is still a major challenge in digital interventions. However, despite the relatively low usage, “ready4life” was shown to be effective [32]. This is consistent with previous research suggesting that users do not necessarily need to complete the entire program to benefit [53], and that after a certain level of use, little additional benefit can be expected [54,55]. Our findings have implications for the design of multi-behavioral digital interventions. The lower usage observed for modules delivered second suggests that more relevant modules (eg, those addressing behaviors with high support needs) should be delivered first. An important finding was that students with higher needs for support in terms of self-efficacy, social competence, and perceived stress showed higher usage. In terms of health equity, additional efforts should be made to increase program usage among males, those with lower levels of education, and those with higher levels of addictive behaviors. A starting point for improving usage could be to pay more attention to the needs and wishes of these groups, for example, by involving them more closely in the development of digital interventions, or by improving usability (eg, language adaptations or supporting explanations for those with lower levels of education).